PET with oxygen-15 water is widely used to determine regional cerebral blood flow (rCBF) changes in functional brain mapping experiments. We and others have shown that when rCBF changes during scan data collection, the measured flow is very close to the flow value at the start of the scan, rather than being a time-average of the flow during the scan. This may have implications for the design of mapping studies which involve two or more components, e.g., a learning or practice phase, and a recall or execution phase. If the phases are performed without attention to when the scan starts, the rCBF maps may reflect a combination of the two phases. If they are coordinated with the scan start, the images may reflect the task phases, permitting one to determine which brain regions are involved in each phase. This hypothesis was tested with a visual memory task which alternated between a 30 sec encoding phase when subjects viewed slides of unfamiliar faces, and a 30 sec recognition phase when they indicated whether they had previously seen faces presented to them. Eleven normal control subjects were studied on the Scanditronix PC2O48-I5B tomograph. Injection of 0-15 water and task onset were coordinated so that subjects performed one of the task phases during the first 20 sec of the scan. The data were analyzed using Statistical Parametric Mapping software. Direct comparison of scans starting in encoding phase with scans starting with recognition showed dramatic differences, with relatively higher rCBF during encoding in the frontal, temporal and parietal regions, more on the left. Comparison of scans starting with each phase to control tasks (a face matching task and a sensorimotor control) demonstrated substantial differences in the patterns of activation. These experimental results were consistent with simulation studies of step changes in rCBF occurring at various times in relation to the start of the scan. Thus both simulations and experimental data demonstrated that the first 15- 20 sec after bolus arrival and scan start determine the observed rCBF response in an activation task. This study provides a strategy for designing functional brain mapping experiments to determine which brain regions are involved in the components of a more complex task.